New drug candidates must be shown not to cause life-threatening Torsades de Pointes arrhythmia. The present metrics used to predict this are block of the hERG potassium current, and prolongation of QT interval. However, some drugs which are strong hERG blockers are not torsadogenic (Kramer, 2013) and not all drugs which prolong QT interval cause Torsades (Segers, 2008). In an effort to find a more specific marker of drug-induced pro-arrhythmic risk, we used multi-ion channel block in combination with mathematical electrophysiological modelling to investigate the link between ion channel effects and susceptibility to afterdepolarisations. Physiological causes of afterdepolarisations include increase in L-type calcium current conductance; decrease in rapid delayed rectifier potassium current conductance; increasing late sodium current (Noble & Noble, 2006); and shifting the voltage inactivation curve of the fast sodium current. We implemented these effects in mathematical models of cardiac cells by altering conductances, concentrations, and ion channel kinetics. The level of these interventions necessary to cause afterdepolarisations were measured (see figure). Using data for drug effects on multiple ion channels, ion channel conductances were reduced to mimic the effects of drug block in the cells, and the intervention thresholds required to provoke afterdepolarisations were measured for each drug. The change in threshold was used to classify drugs into risk categories using linear discriminant analysis, based on training data of clinical drug-induced torsades incidence. The errors in classification were used as a metric for the predictivity of the intervention. All but five of the interventions were more predictive than hERG-only risk markers, which had a mean error of 1.5, with a standard deviation of 1.2. The lowest error was the L-type calcium current increase protocol in the Ten Tusscher 2006 M cell model, with mean error 0.48 and standard deviation 0.62, the same mean error as the APD90 measure developed by Mirams et al. (2011). These results indicate that simulating afterdepolarisation tendency has potential for use in the early stages of drug development as an improved marker for CiPA and drug companies to use for predicting drug-induced arrhythmic risk.